Genetic defects in N-glycosylation and cellular diversity in mammals

Glycoproteins in mammalian cells are modified with complex-type aspargine-linked glycans of variable chain lengths and composition. Observations of mice carrying mutations in glycosyltransferase genes imply that N -glycan structures regulate T-cell receptor clustering and hence sensitivity to agonists. We argue that the heterogeneity inherent in N -glycosylation contributes to cellular diversity and, thereby, to adaptability in the immune system. This paper reviews the type I CDG syndromes, with a focus on the clinical symptoms and the mis-sense mutations in the PMM2 gene. This issue of the journal has several other reviews on CDG syndromes that cover all aspects of the problem. The autoimmune and T-cell hypersensitivity phenotypes in Mgat5 –/– mice are documented. Furthermore, Mgat5 –/– T cells show enhanced TCR clustering, receptor internalization and intracellular signaling. The presence of a glyco- protein–galectin lattice on the cells was demonstrated by competition with sugars and by chemical cross-linking of TCR and galectin-3. The results sug- gest a general mechanism for regulation of receptor clustering through differential glycosylation and interaction with cell-surface lectins.


Introduction
N-and O-linked glycans are found on both cell-surface and secreted proteins, many of which control proliferation and cell fate decisions in animals. Tissue-specific expression of glycosyltransferases is a significant factor controlling the glycan profiles observed in differentiated cells [1]. In addition, many glycosyltransferases compete for acceptor intermediates, causing bifurcations of the pathways and additional structural complexity [2]. Heterogeneity at specific Asn-X-Ser/Thr sites on individual glycoproteins is very common and diversifies the molecular population into glycoforms. The protein environment of each Asn-X-Ser/Thr influences access by processing enzymes and inefficiencies in this process result in only a fraction of N-glycans receiving certain substitutions. As a result, each Asn-X-Ser/Thr site of a glycoprotein is commonly populated by a set of biosynthetically related N-glycan structures.
The specific activity with respect to a particular functional property can vary between individual glycoforms and thus the potency of a glycoprotein is actually the weighted average of the specific activities of the glycoform population. For example, the peptide hormones lutropin and erythropoietin are produced with structurally diverse glycans, and the discrete glycoforms differ in their affinity for hepatic lectins, thereby determining serum half-lives and potency in vivo [3,4]. By balancing the proportions of glycoforms with different serum half-life, the potency and kinetics of a cytokine response can be modified or adapted to extrinsic conditions. In a similar manner, variation in receptor glycosylation has the potential to modify their physical associations in the membrane and thereby the ligand occupancy thresholds for signal transduction. The terminal portions of mature N-glycans are generally not held close to the glycoprotein on which they reside. This leaves them free to bind to multivalent lectins at the cell surface. Various receptor glycoforms would have differing affinities for cell-surface lectins. The interaction of receptors with such lectins could potentially serve as an adjustable means of linking various signaling systems and hence tuning the cellular responses to various stimuli, as previously suggested by Feizi and Childs [5]. Receptor systems that form highly cooperative macromolecular clusters in response to agonists, such as T-cell receptors (TCRs) and cell adhesion receptors, appear to be sensitive to glycoform variation. Herein we discuss phenotypes associated with mutations in enzymes of the glycosylation pathways and their implications concerning the molecular functions of cell-surface N-glycans.

Congenital disorders of glycosylation
The first stage of protein N-glycosylation, conserved from yeast to man, is the synthesis of Glc 3 Man 9 GlcNAc 2 -PPdolichol and transfer of the oligosaccharide to proteins in the ER. Five congenital disorders of glycosylation (CDGs) are known, each with a characteristic enzymatic defect either in the biosynthesis of Glc 3 Man 9 GlcNAc 2 -PPdolichol or in its transfer to asparagine. These are known as type I CDGs. Infants with these defects share clinical features, including developmental delay, multiple organ abnormalities and severe neurologic dysfunction [6 • ,7]. These deficiencies result from the partial failure to glycosylate proteins with oligosaccharides at the usual Asn-X-Ser/Thr sites. A complete deficit of N-glycosylation is expected to be incompatible with life, as the inhibition of oligosaccharyltransferase by tunicamycin is toxic to yeast and mammalian cells. Glc 1 Man 7-9 GlcNAc 2 on newly synthesized glycoproteins binds calnexin, calreticulin and ER α-glucosyltransferase. The last acts as a sensor of glycoprotein conformations and, combined with the action of α-glucosidase II, a deglucosylation/reglucosylation cycle continues and retains the glycoprotein until proper folding occurs, thus fulfilling a function analogous to that of chaperones [8 • ,9]. CDG type I patients display a remarkable variability in clinical phenotypes, even, for example, between patients that have the same mutant alleles of the phosphomannomutase gene PMM2 [6 • ]. The genetic background of each patient can presumably influence the expressivity of a phenotype in a complex manner. The chaperone function of Glc 1 Man 7-9 GlcNAc 2 may vary depending on polymorphism in the underlying polypeptides and this could influence the clinical phenotype. 'Silent' polymorphisms are common in the population, such as the underglycosylated state present in CDG patients, and may be innocuous until revealed by a 'molecular stress'. This is analogous to the inactivation of the chaperone Hsp90 in Drosophila, which revealed the accumulated genetic variation between strain backgrounds as morphological mutations [10]. In an evolutionary context, environmental stresses that compromise the buffering capacity of protein chaperones, including those that rely on N-glycans, may reveal a reservoir of genetic diversity for rapid adaptation [10].

Complex-type N-glycans in embryogenesis and immune regulation
The position and specificity of key enzymes in the N-glycan biosynthesis pathway are depicted in Figure 1a. The GlcNAc-TI gene Mgat1 is ubiquitously expressed and required in the biosynthesis of all hybrid and complextype N-glycans ( Figure 1a). Mgat1 -/mouse embryos die at around embryonic day E9.5 with failures of neural tube formation, vascularization and determination of left/right body plan asymmetry [11,12]. Maternal sources of the Mgat1 enzyme and its N-glycan products are present in pre-E6 day embryos, and may support development until E9.5, after which complex-type glycans are largely depleted [13]. In chimeric embryos, Mgat1 -/cells contribute to various tissues, with the exception of lung bronchial epithelium [14]. Consistent with this observation, the Mgat1 -/embryos display a severe failure to organize bronchial epithelium. As Mgat1 mutant Chinese hamster ovary cells grow normally in culture, the in vivo data on the failure of Mgat1 -/embryos support the notion that complex-type N-glycans mediate cell-cell interactions in the animal. However, when growth conditions are limiting, as in low-serum cultures, mutant tumor cells with defects in either GlcNAc-TV (Mgat5) or the Golgi UDP-Gal transporter were observed to grow at reduced rates and displayed an increased dependency on autocrine growth factors [15].
These studies suggested that complex-type N-glycans on certain cell-surface receptors affect their sensitivity to growth factors and/or adhesion signals. Indeed, some glycosyltransferase mutant mice may appear normal at birth. Phenotypes can be revealed, however, by stress and aging, as well as by expressing the mutations in different inbred strains of mice. In this regard, there was a relatively low incidence of mammary tumor growth and metastases induced by the polyomavirus middle T oncogene (PyMT) in Mgat5 -/mice on the 129/sv x FVB background. The PyMT protein is an intracellular docking protein that transforms cells by activating the Ras signaling pathways, as well as phosphatidylinositol 3-kinase and protein kinase B (PKB/Akt). The activation of these latter enzymes is the downstream effect of focal adhesion signaling, which appears to be impaired in Mgat5 -/tumor cells and embryonic fibroblasts. Glycans modified by Mgat5 are present on integrin α5β1 and their depletion may stabilize substratum attachment and result in an impairment of PyMT signaling and tumor growth in A cancer phenotype has also been revealed in Mgat3 -/mice by treating them with a carcinogen. Diethylnitrosamine-induced hepatocarcinomas progress more slowly in Mgat3 -/mice compared with wild-type littermates. Further analysis of the mice revealed that a paracrine growth factor dependent upon Mgat3 promotes tumor growth in the mice [17 • ].
In the absence of GlcNAc-TII (Mgat2), complex-type N-glycans are replaced by hybrid-type glycans ( Figure 1a). Mis-sense mutations have been identified in Mgat2 that reduce enzyme activity by >95% in two CDG type IIa patients [18]. This is a rare autosomal recessive disorder characterized by multisystemic involvement and severe impairment of the nervous system. Mgat2 -/mice are runted and die at variable times after birth with multiple organ defects (D Chui, JD Marth, personal communication), further evidence that complex-type N-glycans are required for normal embryogenesis.
Mutations disrupting fewer glycan structures generally result in viable animals with tissue-restricted or conditional phenotypes. These include mutations that affect either late steps in the processing pathway or enzymes with functional redundancies. Immune phenotypes have been detected in a number of viable glycosyltransferase mutant mice, indicating a particular sensitivity of immune cells to changes in glycosylation. Sialyltransferase ST6Gal modifies N-linked glycans with α2-6-linked sialic acid; mice lacking this enzyme display impaired B-cell proliferation [19]. Mice deficient in ST3Gal-1, a sialyltransferase that substitutes core 1 (Galβ1-3GalNAc) O-glycans, also display the subtle phenotype of CD8 + T-cell depletion by apoptosis, but retention of CD8 + memory T cells [20]. The E-, Pand L-selectins and their ligands control leukocyte traffic. Mutations in several glycosyltransferases that produce selectin ligands disrupt trafficking, but mutant phenotypes differ qualitatively, presumably because of the details of glycan structure and their tissue distribution. These include Fuc-TIV, Fuc-TVII [21], the O-linked core 2 GlcNAc-T(L) branching enzyme [22] and a β3GlcNAc-T that extends core 1 O-glycans with 6-sulfo sialyl-LeX [23]. Mice deficient in the N-linked processing enzymes α-mannosidase II [24] or Mgat5 both develop autoimmune kidney disease with age [25 •• ].

Mgat5-modified glycans regulate TCR sensitivity to antigen
Mgat5 -/mice lack detectable GlcNAcβ1,6Manα1,6branched N-glycan products and are born healthy from CD1 outbred and 129/sv inbred mouse strains [16 However, small litter sizes and higher perinatal mortality have been observed with the Mgat5 mutation on a PLJ mouse background, a hyperimmune mouse strain (J Pawling, M Demetriou et al., unpublished data). Mgat5 -/mice displayed an age-dependent autoimmune disease characterized by glomerulonephritis [25 •• ]. T-cell-dependent immune reactions were exaggerated in Mgat5 -/mice and T cells in culture were hypersensitive to TCR agonists. Delayed type hypersensitivity was more severe and susceptibility to experimental autoimmune encephalomyelitis, a model for human multiple sclerosis, was greater in mutant mice. In contrast, Mgat5 -/-B cells responded normally to a variety of stimuli, indicating that the defect is restricted to selected cell types.
TCRs are recruited into 'immune synapses' by peptide-MHC complexes on antigen-presenting cells (APCs). A threshold number of receptors are required in the immune synapse to sustain intracellular signaling and trigger entry into S phase [26]. This rate-limiting event of sustained TCR clustering was enhanced in Mgat5 -/cells ( Figure 2). TCR-dependent tyrosine phosphorylation,   Some glycoproteins recycle through the endocytic pathway and back to the cell surface, but this does not appear to involve remodeling of N-glycan branching [27]. As such, changes in N-glycan branching at the cell surface occur through de novo synthesis of glycoproteins. The timescale of this process is hours, whereas TCR-dependent tyrosine phosphorylation, Ca 2+ mobilization and actin rearrangement occur in minutes after agonist-induced receptor clustering. Therefore, Mgat5-dependent regulation of TCR sensitivity to antigen is a long timescale, slow form of negative feedback, governed by the steady-state activity of signaling pathways that are downstream of TCR. This type of delayed negative feedback has been observed in other pathways and provides a mechanism to adjust or tune receptor sensitivity to match ambient conditions [28]. T cells in different tissue environments may alter Mgat5-modified glycan levels, thereby adapting the TCR sensitivity and response threshold.

Lectin-glycan interactions modulate lymphocyte receptor signaling
The TCR α and β chains together have seven N-glycans and CD3 γ and δ each have one chain. Glycosylation is necessary for the assembly of these peptides into a mature TCR complex. The N-glycans protrude ~30 Å from the protein surface and a fraction of these glycans appear, by their positive leukoagglutinin L-PHA reactivity, to be β1,6GlcNAc-branched complex-type structures [29], which are preferentially substituted with poly N-acetyllactosamine 604 Carbohydrates and glycoconjugates  Current Opinion in Structural Biology [30]. Electron density for the chitobiose core (GlcNAc 2 ) of N-glycans is often observed in high-resolution X-ray structures [31 •• ]. However, density for distal N-glycan sequences is generally not visible in X-ray structures, consistent with their mobility and lack of order in the crystal. Biantennary complex-type N-glycans have been modeled onto TCR complex. Based on their size and positions, it has been proposed that N-glycans may limit nonspecific receptor aggregation and thereby also spurious T-cell activation [31 •• ]. The glycans may also align the geometry for TCR binding towards peptide-MHC complexes on the APC [31 •• ].
We recently reported that glycoproteins of the TCR complex bind to galectins at the cell surface, which impedes TCR clustering in response to agonist [25 •• ]. The galectin family is conserved in metazoans and features either one or two carbohydrate-recognition domains (CRDs) per molecule. Galectin-1 and galectin-3 have one CRD and form homodimers with CRDs spaced ~50 Å apart and oriented in opposing directions [32], a feature ideally suited for crosslinking glycoproteins with multiple glycans [33] (Figure 2). Treating wild-type T cells with lactose or lactosamine to dissociate the galectins from their endogenous ligands enhanced TCR clustering and tyrosine phosphorylation in response to agonist, in essence creating a phenocopy of Mgat5 -/- [25 •• ]. Furthermore, cell-surface galectin-3 was found physically associated with TCR complex, an interaction enhanced by the expression of Mgat5 glycans. Monomeric affinities of galectins for lactosamine and lactose are in the 10 -3 M range [34], which is comparable to peptide-MHC-induced oligomerization of TCR measured in solution and lower than TCR affinity for peptide-MHC to form an immune synapse [26]. Poly N-acetyllactosamine, a slightly higher affinity ligand (10 -4 M) for galectin-3 [34], is preferentially added to Mgat5-modified glycans [30] ( Figure 1a). Therefore, Mgat5 glycans on the TCR complex and other glycoproteins appear to form a multivalent lattice held together by galectins, which slow the migration of the TCR into clusters at the immune synapse (Figure 2a,c).
There are at least ten mammalian galectins with overlapping expression patterns. Mice deficient in galectin-1 or galectin-3 are healthy, but have not yet been examined for T cell or stress-related phenotypes [35]. Galectin-3 and the TCR complex are present in the T-cell lattice and galectin-1 has been reported to bind to glycoproteins designated CD2, CD3, CD4, CD7, CD43 and CD45 on T cells [36], but other components of the lattice remain to be defined. Exogenous galectin-3 and galectin-1 modulate T-cell activation in vitro [36], antagonize TCR signaling [37 • ] and, when injected into mice, galectin-1 can suppress autoimmune disease [38]. The expression of galectin-3 at the cell surface of cancer cells is associated with an increase in tumor growth, invasion and metastasis [39 • ], possibly as a result of enhanced turnover of integrin-substratum contacts, which would give rise to an increase in focal adhesions.
Siglecs are sialic-acid-binding lectins implicated in lymphoid and myeloid cell functions. CD22 (siglec-2) binds ST6Gal products (SAα2,6Galβ) in cis on the B-cell surface. Unlike galectins, CD22 is a transmembrane protein with a phosphorylated cytosolic domain, and recruits Grb2, Shc, SHP1 and SHIP, causing reduced B-cell receptor signaling [40]. B cells of CD22-deficient mice are hypersensitive to antigen stimulation. However, ST6Gal-deficient mice display impaired B-cell proliferation, attenuated antibody production and, although cell-surface CD22 is present, it is not bound to ligand [41]. This suggests that the recruitment of CD22 into B-cell receptor signaling complexes is inhibited by SAα2,6Galβ. Therefore, SAα2,6Galβ appears to be a negative regulator of a negative regulator (i.e. CD22) and loss of SAα2,6Galβ allows CD22 to dampen the B-cell response. Such a possible scenario is depicted by the model in Figure 1b (represented by the right panel). ST6Gal activity [42] and CD22 occupancy with sialic acid are regulated with B-cell activation [43]. Similar to Mgat5 regulation in T cells, this suggests that the signaling threshold for B-cell receptors may be regulated by differential sialylation, which controls the availability of CD22.

Structural diversity of glycans increases functional diversity
Individual cells typically show small variances of many molecular parameters that, together, confer a Gaussian spectrum of responsiveness in the cell population. The protein glycosylation machinery appears to be designed to increase molecular heterogeneity and, presumably in some instances, functional diversity within cell populations as well. Although different receptor glycoforms may vary only slightly in their affinities for lectins and signal transduction, exponential amplification of lymphocyte clones can convert small differences into large systemic events [44]. T-cell clones undergo multiple rounds of cell division once triggered by peptide-MHC binding above a threshold affinity. Once this stochastic event occurs, strong positive feedback by cytokines creates a highly cooperative and sustained expansion of cells. Many other interactions between cells influence the balance of Th1/Th2 helper T cells, development of memory T cells and cessation of the response. The immune system appears to be particularly sensitive to small functional differences at the cellular level.
Sustained clustering of ~8000 TCRs results in an increase in the concentration of protein kinases and of docking proteins on the inner surface of the membrane; this results in the triggering of T-cell activation [45]. CD28 is a co-receptor of the TCR. Binding of CD28 to CD80 on APCs enhances the recruitment of intracellular signaling molecules, thereby reducing by fivefold the number of TCRs required for activation [46,47]. The Mgat5 deficiency sensitizes this system still further and lowers the threshold for the TCR response to agonists; this is independent of the ligation of CD28 [25 •• ] (Figure 1c). In addition, the apparent Hill coefficient (n H ) was increased in Mgat5-deficienct cells and this represents the synchrony of the responding cell population in this experiment. Although each T cell responds individually in a switch-like manner, the inherent heterogeneity between cells for N-glycans and other factors such as CD28 levels tends to reduce the synchrony or apparent cooperativity of the responding cell population (Figure 1c) [48]. The Mgat5 mutation eliminates a subset of N-glycans, reducing both molecular heterogeneity in the remaining N-glycans and presumably the heterogeneity of the glycoprotein-galectin lattice (Figure 1b). Mgat5 glycans are present on α1β5 integrin and, in a similar manner, Mgat5 -/cells show increased cooperativity for adhesion [16 •• ], whereas overexpression of Mgat5 has the opposite effect on the adhesion to a substratum of fibronectin [49] (Figure 1d).
Coffey [50] has argued convincingly that complex biological systems, such as the immune system, might be modeled using nonlinear dynamics and chaos theory, beginning with simple rules of interaction governing the behavior of cells in the system. The calculated outcomes over time appear unpredictable, but they are confined to a bounded set and are highly dependent on basal conditions and particularly the genetic background. Coffey further notes that living systems tend to evolve towards the boundary of order and chaos where the system has maximum fitness. Loss of diversity in the lymphocyte population is thought to have an influence on the probability of acquiring infections and autoimmune diseases, as suggested by many gene knockout experiments in mice. For example, CD28-deficient mice are more resistant to autoimmune disease, but more sensitive to certain pathogens. Resistance to infections and development of cancer versus the maintenance of self-tolerance are, in some respects, opposing objectives of the immune system and are balanced for overall fitness. Mgat5deficient mice are prone to developing autoimmune diseases, but they show a relatively low incidence of breast cancer progression in a transgenic oncogene model of cancer [16 •• ].

Conclusions
Structurally diverse glycans are present on many receptors that may also be regulated by interactions with lectins, as discussed here for TCR, integrins and CD22. It would appear that the incremental effects of glycoform variation on highly cooperative events, such as TCR clustering at the molecular level and lymphocyte activation at the cell population level, can set thresholds for systemic events such as susceptibility to autoimmune disease. Although further evidence is needed to support our hypothesis, it is possible that the structural diversity of glycans may be an important feature of adaptability and fitness designed to regulate receptor thresholds and variances of cellular responses to extrinsic stimuli.